Our love affair with AI's large learning models, such as ChatGPT, is already making an impact to our every day lives. Sassy robot butlers seem just over the horizon. With AI, we aim to advance beyond what software currently does best - empowering us, reducing tedious tasks, and making us superhuman. It seems like every day we’re finding new ways to improve the user experience.

We’re also urging caution. New technology brings new ways of solving old problems. The temptation is to focus on the AI part without seeing how it benefits the user. Worse still, is riding the wave and applying AI to mediocre ideas in order to standout. We’ve been here before.

From the user's perspective, we believe there are many opportunities to create ideal user experiences with AI. However, the fundamentals of good software experiences have not changed: empower the user with the tasks they want to do and minimise the ones they either don't want or can't do by themselves.

Good AI in action

While AI isn’t necessarily new, the most effective use is in how it lives under the hood. The AI is there as a considerate friend, a guide, a servant at your beck and call. It’s either helping you or moving things out of your way.

• Composing an email in Gmail allows for it to do a sort of ‘super’ autocomplete, to finish emails off
• Notion AI tool allows for several options to help with writing or generating ideas.
• Adobe is making Photoshop be able to do more with their generative background fill
• Commonwealth Bank, as one example, use AI for pattern discovery and Fraud Detection
• Git Co-Pilot assists developers with code suggestions that’s contextually relevant.

Notion’s AI toolkit makes suggestions for writing. However, I don’t know if it necessarily makes the user into a better writer. The jury is still out whether AI helps writing OR reduces the need to write clearly, but that’s a topic for another day.

A few predictions :

• One big thing we’re already seeing is the ability for unstructured data, like an email or conversation, to be converted to a structured format suitable for an API or some other piece of software. Scenarios where documents need to be read, interpreted and actioned will be ripe for further reducing friction.
• How we describe and label AI-fused actions may not necessarily need to invoke ‘AI’ in the label. I.e. “Remove background” is clearer than “AI Background remover”.
• Navigational tools, like search, can be better used to parse search queries. Instead of ‘Men’s Nikes’ a user could search using sentiment ‘Something sporty and fun’.
• Prompts for generating images, such as Mid-Journey or Dall-E , are still a pain. Further investment into a UI will be needed to move it from early adopters to the general public (though other platforms will most likely adopt it and apply their familiar UI to need the needs of the user. See: Canva)
• JasperAI, a writing tool for marketers, will be vulnerable to users going direct to ChatGPT. They will either pivot or die by the AI-sword.
• AI is very expensive, computationally. Pricing plans that target a per-use basis will become more common as more products embed AI.
• Very long term - Our big bet is AI will be able to navigate software including using authentication to access bank accounts (scenario: ‘hey Siri, transfer $50 to my nephew for my birthday’). I can see Apple introducing Siri 2.0 that is far more capable and, importantly, more trusted to do riskier operations. Apple’s next Trillion will come from mining trust, not data.
• In the realms of AI-made art, taste and curation will become key. Will this be AI-lead? Potentially, but AI isn’t ‘smart’ - it operates on probability. And humans are notoriously hard to peg down when it comes to taste.
• Relationships & trust will still be hard-earned. You cannot outsource trust and connection to AI. That’s what alcohol and coffee is for.

‍